J . Am. Chem. SOC.1993,115, 1095-1 105
1095
Characterization of Chloride Ion Binding to Human Serum Albumin Using C1 NMR Null Point Spectral Analysis William S. Price, Nien-Hui Ge, Luan-Ze Hong, and Lian-Pin Hwang* Contribution from the Department of Chemistry, National Taiwan University, Taipei, Taiwan, Republic of China, and Institute of Atomic and Molecular Sciences, Academia Sinica. Taipei, Taiwan, Republic of China. Received June 4 , 1992
Abstract: CI NMR null point spectra have been applied to study the binding of chloride ions to human serum albumin. The density matrix formalism was used to simulate the null point spectra. We found that (1) the results were consistent with a weak and a strong class of binding sites; (2) relaxation at the weak sites could be modeled using a single correlation time for the fluctuation of the electric field gradient at the bound site, whereas binding at the strong site was modeled using a two-step model; (3) for the weak sites the enthalpy of dissociation was 20.2 kJ mol-', and at 310 K the correlation time was 3.4 ns; and (4) for the strong site at 310 K the slow correlation time was 35 ns and the upper limit for the fast correlation time was 0.5 ns, and the dissociation constant for the strong site was determined to be 0.1 M. This study shows that analysis of the fine spectral structure at the null point provides a more detailed source of motional information than the normally used measurements of longitudinal and transverse relaxation. This method also allows more accurate determination of correlation times and enables the interpretation of quadrupolar ion binding data when the relaxation is too nonexponential to allow the use of usual methods.
1. Introduction
Human serum albumin (HSA) has a molecular weight M Pof 66 439 and is the most abundant protein in blood plasma, with a concentration of about 42 g L-'.' HSA is a single polypeptide chain consisting of 585 amino acid^.^.^ HSA provides about 80% of the colloid osmotic pressure of the blood and plays an important role in the binding and transport of endogenous and exogenous compo~nds.~ In solution, HSA has the shape of a prolate ellipsoid with dimensions 144- X 45- X 2 2 - k 5 Serum albumin exhibits a great deal of homology between different animal species. Extensive reviews of HSA and its properties may be found elsehere.^.^,^ HSA is known to bind a number of quadrupolar ions including CI-, Br-, I-, and A13+.8,9 Chloride ion binding to HSA and bovine serum albumin (BSA) has been studied by a variety of including electrophoresis, equilibrium dialysis, ion-selective electrodes, and NMR.8*1"'9 It has been shown that there is little or no difference between the chloride ion binding properties of HSA and BSA.12J5 The results of previous N M R studies suggest that there are two classes of chloride binding sites on H S A a small number of strong binding sites, to which chloride binding can be inhibited by stoichiometric amounts of sodium dodecyl sulfate (SDS)?,l9 and weak (SDS-insensitive) binding sites. The C1- exchange between protein binding sites and the bulk solution state is believed to be in the fast exchange limit,'6q'9with the free population p f greatly exceeding the bound population P b (at least for the chloride concentrations used to date in N M R experiments). Wl and 37ClN M R provide a convenient means for studying the exchange and binding of chloride ions between free solution and protein binding Previous N M R studies of chloride binding to albumin have been performed at low magnetic field strengths (12.5 T) and have relied upon measuring the excess longitudinal or transverse relaxation rates (Le., the difference between the measured relaxation rate and that measured in the absence of protein) resulting from the chloride ions binding to HSA.8-16'9*22The excess relaxation data were then analyzed to extract the fluctuation correlation time of electronic field gradient at the bound sites T,,. The bound population occurs in the relaxation equations as a product of the square of the quadrupolar coupling constant of the bound sites Xb (i.e., X b = e2Qq/h,where eq is the electric field gradient at the nucleus at the bound site, eQ is the electric quadrupole moment of the nucleus, and h is the Planck constant). This product forms a constant, and the product can only be separated if one of the two terms can be independently determined. However, generally the flucutation correlation time
* Author to whom correspondence should be addressed.
of the electric field gradient and the quadrupole coupling constant at the free site can be independently determined. Chloride ions bound to HSA are outside the extreme narrowing condition (i.e,, UoTb ,*
(dsl9
and P'w
= [(a!),,
( d W 1
To guarantee the validity of eqs 7 and 8, the conditions T~~~ >> T~ and rex,>> 7 , must hold. In our particular case the spectral density functions used in R, and R: defined in Price et al.33are replaced by those derived in the twestep model (see Section 2.2.3 and Appendix 3). 2.2.3. Two-step Model-Free Approach. The internal motion at the strong binding site is accounted for by a modification of the approach of Halle and W e n n e r ~ t r o m . The ~ ~ spectral density functions relevant to this model are given in Appendix 3. Experimentally, it was observed from the fitting of null point spectra that the parameter sets used satisfied the following two relations psx;(1
+ v2/3 - A2)T,r
Cl
P,X,2A2~,, = c2
(94 (9b)
where C1and C2 are constants to be determined and v is the asymmetry parameter. Later it was noted that in the system studied the conditions uOTSf > 1 held (see below) and that in this limit Jlfand JZf can both be replaced by Jorand then JM and Joswill dominate the relaxation process (see Appendix 3). This then causes the relationships expressed in eq 9. 2.3. Simulation of Null Point Spectra. Here we describe the use of the above theory in simulating the free induction decay obtained from the inversion-recovery rf pulse sequence (Le., ? T - T - ~ / ~in- A thec ~ case ) of two-site exchange. Null point spectra are merely those spectra acquired at T values such that the net magnetization is near 0. The equilibrium state multipoles can be calculated from the corresponding density matrix elements (see Appendix 2, eq A.2.1) to be (5) 'I2YhBo = (21+ 1)kT and
In fitting null point spectra resulting from three-site exchange, an initial estimate forp, of 0.003 15 was calculated from eq 3b, the concentrations of chloride ion and HSA used in the null point samples, and the assumption that all of the SDS-sensitive sites were occupied by chloride ions; clearly this value constitutes an upper bound for p,. Initial estimates for x,, A , Tsf, and T,, were then provided, and the constrained nonlinear optimization package of Matlab386 (The Mathworks Inc., MA), which is based on the sequential quadratic programming algorithm, was used to find the set of parameters that fit the peak intensity versus T delay in the inversion-recovery pulse sequence and the line widths at one-half and one-third height of the fully relaxed spectrum. These parameters were then used as a starting point for simulating the null point spectra. The value of xs was given a minor adjustment to make the T values coincide with the null point spectra. The constraints used for assessing the goodness of fit of the null point simulations were the dynamic frequency shift, the time dependence of the fine structure of the null point spectra, and the relative intensities of the null point spectra to that of the fully relaxed spectrum. At the end of the fitting process the parameters obtained gave good fits to (1) the null point line shapes, (2) the longitudinal relaxation data, and (3) the line shape of the fully relaxed spectrum. The fitting procedure for the two-site model was similar. 2.4. Evaluation of Longitudinal Relaxation Measurements. Longitudinal relaxation was simulated using the full density matrix formalism (see above), and the net magnetization derived from this was well described by a single exponential. Simulation of longitudinal relaxation measurements derived from CI-ions involved in two-site exchange was in essence the same as that for evaluating the null point spectra except that the intensity of the spectral line obtained from the term [(oh
( 4 e q
=0
Since we have exchanging species, the individual multipoles must be appropriately population weighted. The effects of the ?T pulse were then calculated (See Appendix 2, eq A.2.2.), and thus immediately after the ?r pulse and including population weighting the state multipoles are
(d5)fCW, = -Pf(w,
